How much of the weight lost on a GLP-1 is lean mass

The worry about muscle on GLP-1 drugs rests on a handful of studies that put a subgroup of trial participants through a DXA scan, the low-dose X-ray that splits the body into fat, lean tissue and bone. The numbers are real. They are also smaller studies than the headline trials, and they measure lean mass, which is not the same thing as muscle. This page sets out what each study found and where the gaps are. The practical side, protein and training, is on the muscle loss page.

The two DXA substudies

STEP 1, semaglutide 2.4 mg. Of the 1,961 adults in STEP 1, 140 had DXA scans at the start and at week 68: 95 on semaglutide and 45 on placebo. In that subgroup, weight fell 15.0 percent on semaglutide and 3.6 percent on placebo. On semaglutide, total fat mass fell 19.3 percent, visceral fat (the fat around the organs) fell 27.4 percent, and total lean mass fell 9.7 percent. Because fat fell faster than lean tissue, lean mass as a share of the body rose by 3.0 percentage points. The results were published as a conference abstract in the Journal of the Endocrine Society in 2021, not as a full paper, and the abstract reports only that placebo showed "no major changes".

The figure often quoted, that around 40 percent of the weight lost was lean mass, is not stated in the abstract. It is a calculation. At baseline the semaglutide group was 53.9 percent lean and 43.4 percent fat. A 9.7 percent fall in lean mass and a 19.3 percent fall in fat mass work out to roughly 5.2 and 8.4 percent of starting body weight, so lean tissue was close to 40 percent of the fat plus lean lost. That is an estimate from published averages, and it should be read as one.

SURMOUNT-1, tirzepatide. Of the 2,539 adults in SURMOUNT-1, 160 had DXA scans at baseline and week 72: 124 on tirzepatide (all three doses pooled) and 36 on placebo. Fat mass fell 33.9 percent on tirzepatide against 8.2 percent on placebo. Lean mass fell 10.9 percent against 2.6 percent. Of the weight lost, about 75 percent was fat and 25 percent lean in both groups; the results section gives 74 and 26 percent for tirzepatide. The authors' reading was that the drug did not take a disproportionate share of lean tissue compared with weight loss on placebo. The paper was published in full in Diabetes, Obesity and Metabolism in 2025.

Lean mass is not muscle

DXA lean mass is everything that is not fat or bone: skeletal muscle, but also organs, water, and the glycogen stored with water in muscle and liver. Early in any diet, glycogen and water go first, which inflates the lean figure. Skeletal muscle is usually estimated at around half of lean mass, though that ratio is itself an assumption. So a 9.7 percent fall in lean mass does not mean a 9.7 percent fall in muscle. It does not mean zero either.

The proportion is also not unique to the drugs. In a 2024 JAMA viewpoint, Conte, Hall and Klein argued that the lean share of weight lost on GLP-1 drugs is broadly in line with other ways of losing weight, that people with obesity start with more muscle than average, and that the quality of that muscle may improve as fat inside it falls. A published reply argued the reassurance was premature, because skeletal muscle had not been measured directly in these trials. Both points stand. The data to settle the question in older adults, the group with the least muscle to spare, do not yet exist.

Mass versus strength and function

Neither DXA substudy measured strength. They measured composition. The closest thing to a function result in the headline trials is STEP 1's patient-reported physical functioning score, which improved more on semaglutide than on placebo. That is a questionnaire, not a grip test or a chair-rise test, and it improved while lean mass was falling, which is consistent with carrying less weight being easier on the body. What the trials cannot tell you is whether a given person kept or lost strength. That is measurable at home: the weights you lift, the stairs you climb.

Exercise and protein: what the evidence supports

The best randomised evidence on exercise with a GLP-1 drug is older and used liraglutide. In a Danish trial published in the New England Journal of Medicine in 2021, 195 adults lost an average of 13.1 kg on an eight-week low-calorie diet and were then randomised for a year to exercise, liraglutide 3.0 mg, both, or neither. Compared with placebo, the exercise group kept off 4.1 kg more, the liraglutide group 6.8 kg more and the combination group 9.5 kg more. Body fat percentage fell 3.9 points with the combination, about twice as much as with exercise alone (1.7 points) or liraglutide alone (1.9 points). Only the combination improved HbA1c, insulin sensitivity and cardiorespiratory fitness. The exercise programme was moderate to vigorous training, not a resistance-only protocol, and the trial did not use semaglutide or tirzepatide.

No large trial has yet randomised people on semaglutide or tirzepatide to different protein intakes. The targets in circulation come from expert review. A 2025 review in Obesity Pillars recommends protein above 1.2 g per kilogram of body weight per day, spread across meals, combined with aerobic activity and structured resistance training. That is a reasonable, low-risk default. It is not a trial result.

Muscle-sparing drugs: the BELIEVE trial

Bimagrumab is an antibody that blocks activin type II receptors, a signal that limits muscle growth. In BELIEVE, a phase 2 trial published in Nature Medicine in 2026, 507 adults with obesity were randomised to nine arms for 48 weeks: placebo, two doses of bimagrumab, two doses of semaglutide, and four combinations. Semaglutide was given open label. On semaglutide 2.4 mg alone, 71 percent of the weight lost was fat. With bimagrumab 30 mg/kg added, total loss rose to 17.8 kg and 92 percent of it was fat, with lean mass down only 1.3 kg. Bimagrumab alone produced fat loss with a small gain in lean mass.

The caveats are the authors' own: the p values were not adjusted for multiple comparisons, the results should not be read as definitive treatment effects, and it was phase 2. Bimagrumab is not approved for any use. Phase 3 trials are needed before anyone knows whether keeping lean mass on the scan translates into better strength, function or weight maintenance.

What this adds up to

• On semaglutide and tirzepatide, roughly a quarter to two fifths of weight lost in the trial substudies was lean tissue. Placebo weight loss had a similar split in SURMOUNT-1.
• Lean mass overstates muscle loss. By how much is not settled.
• No trial has shown the drugs reduce strength or function. None was set up to test it properly.
• Exercise plus drug beat either alone for body fat and fitness in the one good randomised trial. Protein targets are expert opinion.

Related: the STEP trials, the SURMOUNT trials, the safety evidence and real-world results.

Logging it

The scale cannot tell fat from muscle, so log the things that can. Protein each day against your target, resistance sessions each week, and the weights you lift, next to the dose and the weight trend. Peptide AI keeps the dose, food, training, weight and any body composition or lab results on one timeline, and the logging is free. The free GLP-1 tracker page shows the setup.